This research project will pioneer the use of mobile wireless sensor networks of Unmanned Aerial Vehicles (UAVs), to sense traffic conditions and roadway perturbations following a disruption event. Traffic conditions are commonly sensed using either fixed sensors, or crowdsourced data. Such measurement data is usually sparse, and, as a result, cannot be used directly to generate usable traffic maps. Such maps are generated using a combination of both traffic sensor measurement data and past information about usual traffic patterns (for example congestion patterns, or historical travel time information). These maps are therefore accurate for most situations, except in severe disruption events. An inexpensive option for improving current traffic monitoring systems is to have mobile sensors, for example a swarm of UAVs, which can obtain additional data on disruptions and their impacts when needed. This award supports research on the theoretical foundations for implementing and operating such a system. The optimal sensor placement problem solved by this research will allow the system to automatically compute the best path that each UAV should take to sense the traffic conditions, enabling quick updates on the traffic situation. This research will benefit the U.S. economy by providing an inexpensive means to sense traffic, on demand, for disruption scenarios, without the need and the cost to deploy additional fixed traffic sensors. The multi-disciplinary approach will help positively impact engineering education and broaden participation of underrepresented persons.
The optimal mobile sensor placement problem in traffic flow is critical to enable efficient traffic monitoring during disruptions events. In such events, the disruptions or capacity losses in the transportation network are not known beforehand, and can be estimated from traffic measurements generated by mobile sensors (for example UAVs). The fundamental question addressed through this project will be: given prior information on likely network disruptions, and possibly given traffic flow sensor data (which can include crowdsourced data), how can a swarm of UAVs carrying traffic sensors over the transportation network be directed to minimize uncertainty in traffic state estimates and improve situational awareness over some time horizon? Addressing this question requires the simultaneous solution of the optimal placement and traffic state estimation problem, and will open new horizons for mobile sensing systems more generally. The research team will develop an efficient forward simulation framework for networks, based on a first order model of traffic flow. Building on this framework, the team will pose the problem of optimally routing a set of UAVs (with kinematic constraints) over the transportation network to minimize the residual uncertainty of traffic state estimation over a finite time horizon, while simultaneously estimating the current state of traffic. The team will also investigate the problem of optimal routing with partial traffic sensor information.
